#linuxcnc-devel Logs

Apr 25 2017

#linuxcnc-devel Calendar

12:32 PM beginner: hello
12:33 PM skunkworks: hello!
12:33 PM beginner: i want to install linuxcnc 2.8 on ubuntu 16.04 so...how do i proceed
12:33 PM beginner: ?
12:33 PM beginner: i searched for a deb file and didnt found one
12:34 PM beginner: i searched for sources...didnt found those either
12:36 PM beginner: am i going blind here?....or am i missing something?....install from bundled cd with an entire os is not an option
12:38 PM beginner: on the linuxcnc irc channel i get only arrogance....and even that from one user that bestowed his....kind omnipotence to...give me his attention
12:38 PM beginner: by telling me to dedicate a computer only for linuxcnc software and install it from the bundle cd with an entire os
12:39 PM beginner: so...how do i proceed?
12:39 PM beginner: please help?...real help?
12:40 PM beginner: skunkworks?
12:40 PM beginner: any ideas?
12:41 PM cradek: do you want to control machinery with it, or just run the simulator?
12:41 PM skunkworks: beginner, maybe this? http://forum.linuxcnc.org/9-installing-linuxcnc/32193-complete-how-to-linux-mint-18-1-32-64-rtai-or-preempt-with-rip-install-or-deb?start=220#91698
12:43 PM beginner: it doesnt matter cradek...i allready installed the low latency kernel beacause i need it for developing audio hardware anyway on the machine
12:43 PM beginner: if i need to install the preemtive RT i can do that too
12:44 PM pcw_mesa: If you are a "beginner" and want to control a machine its going to be much easier to just install the standard wheezy ISO image
12:44 PM pcw_mesa: besides there are still some (minor) unresolved issues with Ubuntu 16.04 / Mint 18
12:45 PM cradek: yes, it matters
12:45 PM beginner: if that can get me any help from you guys.... pretend it's just for simulation...allthough i will control a step four basic 540 with it in the end
12:46 PM beginner: so cradek...how do i get the software installed on my ubuntu 16.04 machine?
12:47 PM skunkworks: beginner, the thread I linked isn't any help? it says the directions worked great for installing in 16.04
12:48 PM beginner: it guides me to the last page...i am backtracking the thread now
12:54 PM beginner: ok the first operation would be to git clone it
12:54 PM beginner: but if i insert this command in terminal git clone git://git.linuxcnc.org/git/linuxcnc.git linuxcnc-dev
12:54 PM beginner: will it give me the last /current version or some older one?
12:55 PM beginner: 2.8 pre seems to be declared stable
12:55 PM beginner: and somehow polished
12:56 PM skunkworks: You get the repository. At that that point you check out which branch you want. (if you don
12:57 PM skunkworks: (if you don't checkout a branch - you get master (development)
12:58 PM skunkworks: http://wiki.linuxcnc.org/cgi-bin/wiki.pl?Git
12:58 PM skunkworks: see - working with branches.
12:59 PM beginner: i am new with linux it is hard to work even in the graphical environment....the idea is that at the final point i want to buiild a deb file that conrtains all the dependencies and instaall it on any fresh install of ubuntu 16.04
12:59 PM beginner: on any pc
01:00 PM beginner: how do you check out that ? will a graphical menu apear or do i have to type it ilke an expert wizard in the terminal ?
01:01 PM beginner: beaacause i am not an expert wizard :|
01:04 PM beginner: ok...i got a lot more help from here than the other channel...way lot more help...atleast now i can get the sources on the pc
01:06 PM beginner: skunkworks do i have to give information about me to get the sources?
01:07 PM beginner: aah...soory didnt see anonymous checkouts
01:07 PM skunkworks: it scans your computer for bootleg media - other than that - no.
01:09 PM seb_kuzminsky: beginner: this might help: http://linuxcnc.org/docs/devel/html/code/building-linuxcnc.html
01:11 PM beginner: thanks guys...for now atleast :)...i will get the sources and after that i will come here for further assistance if something goes horribly wrong
01:14 PM beginner: it is better to have something horribly worng in the installation process than when the macchine operates...beacuse then..you know...drill bits fly around and...freddie krueger and stuff :))
01:15 PM beginner: thank you for now
02:24 PM guru_florida: Hi guys. Running latest 2.7 on wheezy getting a few "task: main loop took 0.125 seconds" messages. Should I be worried? I saw some IRC notes about it. I am running off SSD SATA. I also have a PCIE Mesa card. Computer is a i7 Asus x79 board (latency tests stable at 15 micros jitter max)
02:24 PM cradek: no, that's fine to ignore
02:24 PM guru_florida: I also have a VFD running modbus with python (my own design)
02:24 PM guru_florida: cool, thanks
02:26 PM guru_florida: I also get 'thud' during jogging here and there. Any axis. Doesnt seem to happen during program motion.
02:30 PM cradek: any errors in the gui?
02:30 PM guru_florida: no, not at all.
02:31 PM guru_florida: it kind of seems like a misstep or skip.
02:33 PM guru_florida: Like I say, I cant hear it during programmed motion and my parts dont show signs of missed steps. Though that was on my old setup - a motherboard+cpu circa 2004 lol, I just upgraded and figured it might go away. I dont seem to have SMI problems as latency tests are ok...but I could do more investigation in the SMI area.
02:33 PM guru_florida: bear in mind both old and new setups used Mesa (7i43 and now 6i24) so realtime is not as much an issue then.
02:33 PM cradek: if you aren't getting the error about realtime delay, I doubt you have smi problems
02:34 PM cradek: when exactly does it thud?
02:38 PM guru_florida: When jogging with keyboard. Could be as simply as a key read issue or something. I can investigate more, just thought it might ring a bell with someone. I have yet to cut parts with the new setup so if there are missed steps it will appear.
02:38 PM cradek: what part of the jog? beginning, end, middle?
02:40 PM bpuk: zlog
02:43 PM guru_florida: middle
02:43 PM cradek: every jog, or long jogs only, or at random?
02:44 PM guru_florida: long jogs
02:44 PM cradek: hm, every long jog?
02:45 PM guru_florida: no probably not, but easy to reproduce - not intermittent. seems random positions.
02:45 PM cradek: could you capture it in halscope? you could plot velocity, acceleration, and servo cycle time
02:48 PM guru_florida: sure, I can do that when I get home.
02:48 PM guru_florida: thanks!
02:49 PM bpuk: guru_florida: as a daft question - I'm assuming this is a stepper machine? what drives are you using? (the thud is ringing a vague bell for me)
02:56 PM skunkworks: bpuk, welcome back :)
02:56 PM bpuk: thanks skunkworks - life got busy :)
02:56 PM bpuk: on a totally different note - I've got a N3000 based board that I'm trying to get to behave with PREEMPT_RT. Getting 320-400 ms latency under 4.9.20-rt16, 10-12 us under RTAI.
02:57 PM bpuk: If I can't get this board to work with preempt - anyone got a recommendation for a good current board? (it's being hooked up to a 7i80)
02:57 PM skunkworks: how are you building the rt_preempt?
02:58 PM bpuk: pull the kernel tarball, pull the patch, patch the kernel, change to Fully Preemtable, strip out anything obvious I don't care about
02:58 PM bpuk: then make
02:59 PM skunkworks: ok
02:59 PM skunkworks: (wanted to make sure you are using fully preemptable..)
02:59 PM KGB-linuxcnc: 03andy pugh 05master 5ab9bb9 06linuxcnc 10src/hal/drivers/mesa-hostmot2/sserial.c Remove a compiler warning related to type-punning (and fix the indentation) * 14http://git.linuxcnc.org/?p=linuxcnc.git;a=commitdiff;h=5ab9bb9
03:00 PM bpuk: yeah, I figured that was the question - currently trying a 4.4.39-rt50 build in case a recent patch has broken something, but it's not the fastest machine to build on
03:00 PM skunkworks: my thought would be to try a 3.something kernel for grins
03:01 PM jepler: andypugh: I am worried about that change
03:01 PM jepler: I think it copies 8 times too much data
03:01 PM guru_florida: mpuk: using gecko G203V (yes, steppers)
03:01 PM skunkworks: I have had pretty good luck so far with 4.9
03:01 PM jepler: DataLength is in bits, but memcpy's argument is in bytes
03:01 PM jepler: is my interpretation based on this snippet
03:01 PM andypugh: Good point.
03:02 PM andypugh: Well spotted.
03:02 PM guru_florida: I would suspect the G203s coming out of standby current but thud is in the middle of travel. Maybe thud is not the right word, it's not that loud but not normal either.
03:02 PM andypugh: (ANd thanks for looking)
03:04 PM bpuk: guru_florida: ok, this is a _real_ long shot - but when I had the thud I was running gecko drives (G251's in a G540). The solution for me was inverting the step output
03:04 PM bpuk: guru_florida: some, probably not all, of the gecko drives really care about whether the signal is active high, or active low
03:05 PM guru_florida: wow! ok. :) I will try that.
03:06 PM bpuk: skunkworks: I've tried three host boards on 4.9 - all of which work well under rtai, the worst case I've seen was 3.8 s of jitter
03:06 PM cradek: ooh yes if you have the wrong active levels, the timings can be violated
03:07 PM skunkworks: bpuk, yeck
03:08 PM KGB-linuxcnc: 03andy pugh 05master a1a347d 06linuxcnc 10src/hal/drivers/mesa-hostmot2/sserial.c Remove a potentially disasterous confusion between bits and bytes. * 14http://git.linuxcnc.org/?p=linuxcnc.git;a=commitdiff;h=a1a347d
03:09 PM skunkworks: bpuk, maybe pcw_mesa has some tricks? Have you disabled any power saving settings in the bios?
03:09 PM skunkworks: my laptop.. http://electronicsam.com/images/KandT/testing/Screenshot-63.png
03:10 PM bpuk: skunkworks: disabled everything I could find in the BIOS, set kernel parameters idle=poll, max_cstate=0 (or something like that), few others
03:10 PM bpuk: skunkworks: and yes, that was the picture that made me go - oh, looks like preempt is getting good results! :P
03:11 PM bpuk: guru_florida: The line I used a lot was 'setp hm2_5i25.0.gpio.006.invert_output true' (replace 006 with the pin number of your step output)
03:12 PM skunkworks: bpuk, with the 7i80 - you can have latency approaching 100us or a bit more
03:12 PM skunkworks: So don't try for 20us - I have found 70us more common
03:13 PM bpuk: skunkworks: yup, but when it gets into 10+ms it's a couple of orders of magnitude out
03:13 PM skunkworks: sure
03:13 PM bpuk: I'd be over the moon with 50
03:13 PM skunkworks: :) good - you have the right expectations then
03:14 PM bpuk: hence the - 'if this doesn't work, can anyone suggest a specific board they've had good results with' :)
03:14 PM pcw_mesa: 10+ms sounds like power management still on
03:16 PM skunkworks: I am using the Gigabyte j1900 boards with good results
03:16 PM skunkworks: and 1800
03:18 PM bpuk: pcw_mesa: I've passed 'idle=poll processor.max_c_state=1 intel_idle.max_cstate=0' to the kernel - running powertop indicates it's at C0 100% of the time. I can't confirm the GPU isn't entering a power saving state, but I'm fairly sure that the cpu isn't any more
03:18 PM bpuk: pcw_mesa: but yeah, that was my first thought too
03:19 PM bpuk: interestingly - opening powertop consistently produces a latency spike of 200+ ms
03:19 PM pcw_mesa: I have a 3050 based CPU that has (barely) acceptable latency say 200 usec so will run a 7I80 at 1 Khz
03:19 PM skunkworks: that almost sounds like video
03:21 PM pcw_mesa: Best I have for Preempt-RT is a H97/G3258 which is about 7usec after almost a day of videos/GLXgears
03:21 PM pcw_mesa: (it also will run hm2_eth reliably at 4 KHz)
03:21 PM bpuk: skunkworks: Interesting - I'd scrapped the idea of using a J1900N-D3V based on the OSADL latency charts
03:22 PM bpuk: that... sounds pretty nice - which motherboard is that?
03:22 PM skunkworks: bpuk, it isn't stellar - but <80us (j1800) works great on the matsuura :)
03:27 PM bpuk: I'll see how 4.4 runs tomorrow - if its still unacceptable I'll order a new board then
03:27 PM bpuk: thank you all for the help
03:28 PM guru_florida: I am happy with my mobo results but looking at the N3000 boards with NVME drive sockets and a compact flash would look sweet if we had a Compact Flash Mesa card. :)
03:29 PM bpuk: pcw_mesa: to be honest - if the H97 board you're using is still available, I'm half tempted to order one now
03:30 PM guru_florida: I've been wondering if a RPi3 (quad core ARM64) and an FPGA could be a good LinuxCNC controller in headless mode (remote UI). I'm not the first tho. People are struggling with RT kernel on ARM apparently. An all-in-one board would be nice.
03:30 PM bpuk: on G71 - I haven't forgotten - I'm setting up a second machine as a linuxcnc dev box (instead of trying to code in a VM) and continuing on
03:53 PM linuxcnc-build: build #4933 of 0000.checkin is complete: Failure [4failed] Build details are at http://buildbot.linuxcnc.org/buildbot/builders/0000.checkin/builds/4933 blamelist: andy pugh <andy@bodgesoc.org>
04:22 PM andypugh: guru_florida: Gene on the mailing list has an RPi and a 7i90 running SPI comms. He is having a hard time with it.
04:23 PM seb_kuzminsky: jepler set up an Odroid U3 with SPI to a 7i90 (i think), it needed kernel patches and level-shifting hardware to work well
04:25 PM jepler: Yeah there's no ARM board I can recommend for running LinuxCNC, even though we do have some basic drivers for them.
04:27 PM jepler: The machinekit fork is a lot more strongly aligned with ARM, they offer downloadable OS images for a couple of boards even. If you are interested in ARM for ARM's sake, that's probably the closer path to something that works. Just don't ask us about it here.
04:31 PM seb_kuzminsky: i made some progress on my 'hm2 on the beaglebone pru' project last week, but it's not ready for use yet
04:32 PM seb_kuzminsky: i turned some steppers, with the steps coming out of the pru, whee
04:33 PM seb_kuzminsky: if you're into 4 year old hardware
04:53 PM andypugh: I am quite into 100-year-old hardware
05:11 PM bpuk: well - 4.4.39-rt50 gives 180ms jitter, followed by a hard hang
05:11 PM bpuk: I think it's safe to go with 'this board ain't going to be suitable for this machine'
05:12 PM seb_kuzminsky: good hardware is hard to find
05:12 PM bpuk: So it's a pick between a J1900, a N3150M-E (based on osadl figures), or the H97/G3258 combo mentioned above
05:16 PM seb_kuzminsky: at one point we had a wiki page of hardware test results, but i don't know if it's up to date
05:16 PM bpuk: it's not :( - and doesn't differentiate between RTAPI and PREEMPT_RT
05:16 PM bpuk: I'll be adding my results after a few more tests on this board
05:17 PM seb_kuzminsky: we need some otaku to obsessively monitor the mailing lists, forums, and irc, and edit it all down to a single helpful webpage
05:18 PM bpuk: I'm also kinda curious as to how closely cyclictest compares with latency-test
05:18 PM bpuk: because if it's a good test a Supermicro X9SCM/E31220 looks pretty good
05:20 PM KGB-linuxcnc: 03andypugh 05master b215611 06linuxcnc 10src/emc/usr_intf/stepconf/pages.py Stepconf: Allow users to add MDI_COMMANDS * 14http://git.linuxcnc.org/?p=linuxcnc.git;a=commitdiff;h=b215611
05:26 PM skunkworks_: I had some hp machines that didn't run uspace very well unitil either I turned off the right thing in the bios or 4.9 preempt
05:31 PM bpuk: in fairness to the board I'm currently testing - it's an asrock beebox - intended as a media centre
05:31 PM bpuk: but was suprised it did so well under rtai, and so badly under preempt
07:46 PM Tom_L: i haven't found HP machines to be good at much of anything
08:07 PM skunkworks_: we have had good luck with some HP / compact 2nd gen i5's
08:07 PM skunkworks_: we get them refurbished cheap with windows 7 (for work)
08:08 PM skunkworks_: we have probably bought 30 of them and I don't think we have had one hardware issue
08:08 PM skunkworks_: we
08:08 PM skunkworks_: we
08:08 PM skunkworks_: we
08:09 PM skunkworks_: small form factor with display port + vga. (so they do 2 monitors right out of the box)
08:09 PM skunkworks_: and for basic business work - they are fine